Signal conversion devices are characterized by a frequency response, consisting of amplitude and phase terms. Correction (equalization) of the frequency response is essential for a high quality converter and requires precise measurement of device properties, and the group delay dependence on frequency in particular.
Analog to digital converters (ADCs) operate in a wide frequency region that extends from baseband frequencies up to the limiting case of high frequencies. It means that group delay of an ADC should be measured in the correspondent wide frequency range. To ensure a precise correction of the frequency response of an ADC or a frequency converter the group delay measurement should be performed with a small enough frequency step: the distance between the adjacent frequencies, at which the group delay is measured should be reasonably small-sized. The apparatus for group delay measurement should be straight forward: it should comprise as few components as possible, and should not require employment of laboratory devices and additional calibration.
Group delay measurements are conventionally performed using Vector Network Analyzers (VNAs). For example, Agilent Application Note 5965-7707E “Understanding the Fundamental Principles of Vector Network Analysis” describes a group delay measurement method which injects a known sinusoidal excitation to an input of the device under test (DUT), and analyzes the phase of the signal at its output. However, that VNA-based method has limitations. First, it requires a DUT to have input and output ports of the same type, As a consequence, that method is not applicable to such devices as ADCs, where the input signal is analog and the output signal is digital. Second, VNA-based method of measurement requires that the input and the output signals lie in the same frequency range. For these reasons, VNA-based group delay measurement cannot be used for frequency converters.
Another prior art method for group delay measurement is based on a time domain pulse shape analysis. The time domain method uses an injection of a known periodic signal (e.g., impulses or rectangular pulses) into a DUT and capturing a waveform at the DUT output. This method is applicable to ADC group delay measurement. Phase distortions introduced by the DUT can be determined using Fourier Transform-based analysis, i.e., obtaining a spectrum (and more generally, the phase response) of the DUT output signal and comparing it with the spectrum (or the phase response) of the input signal, thus obtaining the phase response of the DUT. However, high frequency measurements require expensive and complicated tools for test signal generation, such as a special picosecond pulse source. Signal sources of that kind may have varying group delay, for which reason they must be calibrated before the measurement, using, for example, a high accuracy temporal resolution sampling scope. As a consequence, the measuring apparatus becomes complicated and cumbersome, ruling out the possibility of calibrating an ADC (or like devices) under operating conditions.
European Patent Application No. EP1515147A1 (J. Kraus and C. Kikkert) describes group delay measurement of an ADC based on generating a test signal through modulation of an initial signal consisting of a plurality of spectral components, by a low frequency signal (FIG. 1). With that method, a digitized waveform from the output of a DUT is Fourier transformed. The phase differences between sidebands of initial signal spectral lines are calculated. Group delay at the corresponding frequency is found by dividing the corresponding phase difference by the difference of the sideband frequencies.
A phase difference between the sidebands of the initial signal spectral line contains an unknown phase offset that equals a doubled phase of the low frequency signal. This phase offset causes a corresponding offset in the calculated group delay. As long as the low frequency signal remains unchanged during the measurement, the unknown offset delay is inconsequential—it does not cause distortions of the measured group delay.
The number of spectral lines in the initial signal is limited by the acceptable complexity of the measuring apparatus, and usually is far less than the number of frequencies where group delay is to be measured. A repetition of the measurement, with alteration of the spectral lines frequencies in the initial signal, presents difficulties because it is accompanied by random changes in the phase of the low frequency signal with the a corresponding appearance of the unknown offset in the measured group delay. As a consequence of the limited number of frequencies where group delay is measured, only an approximate correction of frequency response based on these measurements is possible. This factor significantly constrains applicability of this measurement method.
U.S. Pat. No. 8,983,796 (T. Bednorz and S. Neidhardt) describes a different method of group delay measurement based on generation of two sine wave signals with different frequencies f1 and f2. This method is schematically illustrated in FIG. 2. The input signals f1, f2 from two sine wave signal generators 10 and 11 are added in a summer 12. The resultant summed signal is split in a splitter 13 between a reference channel 40 and a DUT 20 followed by a measurement channel 30.
The phase difference of the input signals is measured in reference channel 40 using a mixer 41 which uses a local oscillator signal fLO to translate the summed signal from splitter 13 to a low frequency signal including components derived from input signals f1, f2. That two component low frequency signal is processed to perform phase detection using a digital quadrature circuit consisting of low pass filters (LPFs) 42 and 43 followed by analog to digital converters (ADCs) 44 and 45 respectively, followed by a phase detector 60. In this configuration, the outputs of ADCs 44 and 45 are input to phase detector 60, which may be implemented as quadrature phase detector.
The phase difference of the output of the DUT is measured in measurement channel 30 followed by a phase detector 50, which are similar to reference channel 40 and phase detector 50. In particular, measurement channel 30 includes a mixer 31 which uses a local oscillator signal fLO to translate the summed signal from splitter 13 and the DUT 20 to a low frequency signal including components derived from input signals f1, f2. That two component low frequency signal is processed to perform phase detection using a digital quadrature circuit consisting of low pass filters (LPFs) 32 and 33 followed by analog to digital converters (ADCs) 34 and 35 respectively, followed by phase detector 50. In this configuration, the outputs of ADCs 34 and 35 are input to phase detector 50, which may be implemented as quadrature phase detector.
The frequencies f1, f2 of the input sine wave signals are swept within a frequency band of interest and group delay measurement is obtained by subtracting the reference phase difference (at the output of phase detector 60) from DUT 20 phase difference (at the output of phase detector 60) in a group delay calculation unit 70. This method also has a number of disadvantages and suffers from circuit complexity. It is not possible to measure group delay of an ADC at low frequencies. Moreover, there is a need for wideband mixers 31 and 41 which are used for providing signal down-conversion. The down-conversion step includes use of low pass filters 32, 33, 42 and 43, which may introduce additional phase distortions and so must be precisely matched for four quadrature branches (reference and measurement). Also, each quadrature channel is digitized by separate ADCs 34, 35, 44 and 45, which may have frequency mismatch and, as a consequence, introduce phase errors.
There is a need in a method and a simple apparatus for group delay measurement which may be applied to signal conversion devices and which provide for precise measurement of group delay in a wide frequency band at frequently repeated frequencies by separate tests for different sets of frequencies.